Neurocognitive Impairment in Non-Central Nervous System Cancer Survivors
Exploratory Study of Attention and Executive Functioning Assessments Using Event-Related Potentials
Abstract
Abstract: Cancer and its treatments entail a profound inflammatory response of the central nervous system (CNS). This intense neurotoxic process can lead to significant neurocognitive impairment even in non-CNS cancers. Few studies have examined this domain, and data available is based on limited designs using neuropsychological assessments comprising self-report or traditional testing batteries that capture basic response data. Here, we leverage cognitive electrophysiology, specifically Event-Related Potentials (ERPs), to examine and delineate neurocognitive impairments in non-CNS cancer survivors. Eleven survivors, who were on average 4.6 years in remission from head and neck cancer and 10 matched healthy controls underwent standardized cognitive and emotional “Go-Nogo” paradigms concomitant to EEG recording. Significant differences in amplitude morphology in the very early ERP components (C1, N1, P1) and middle ERP component (N2), were apparent between non-CNS cancer survivors and controls. Later ERP components (P3, N4) did not show amplitude differences. Non-CNS cancer survivors yielded faster latencies in the early components for pain-related stimuli during the emotional paradigm, albeit tended to yield slower ERP latencies overall across both experiments. These findings suggest that early gating and inhibitory control are dysregulated in non-CNS cancer survivors, which can impact executive functioning and cognitive processing involved in everyday activities for many years post-treatment. The findings do not fully align with ERP morphologies associated with neurocognitive impairment in degenerative conditions (such as dementia and Alzheimer’s disease), affecting later-stage ERP components. Our initial results suggest that (1) cognitive impairments in cancer survivors do not follow a degenerative trajectory, (2) rather are in line with “lesion” related damage (such as stroke, epilepsy), and as such, (3) have the potential for treatments involving neurocognitive plasticity.
Accruing evidence supports the concept that non-central nervous system (CNS) cancers and their treatments ensure significant damage/injury to the central nervous system (Cerulla Torrente et al., 2020; Hervey-Jumper & Monje, 2021). Centralized nervous system compromise is attributed to significant impairments in cognitive functioning. Various pathophysiological processes including proinflammatory effects within the cancer microenvironment, in addition to neurotoxic effects of cancer treatments to neurons, glia, and cortical-subcortical brain region connectivity, have been ascribed to neurocognitive symptomatology in survivors across discrete non-CNS cancers (Hervey-Jumper & Monje, 2021). This highlights the significance of recognizing and classifying impairments in neurocognitive domains in cancer survivors to ensure successful management and quality of life.
Presently the tools used or available might not be optimal for this task. Studies examining neurocognitive impairments in non-CNS cancer survivors have largely focused on detriments to cognitive functioning associated with radiation and/or chemoradiation treatments, the main treatments received. Albeit, even newer treatments involving targeted therapy, hormone therapy, and immunotherapy also appear to affect cognitive functioning in cancer populations at large (Amidi et al., 2017). Translating these emerging patterns in terms of neurocognitive impairment is difficult however, since the data available is based on either (1) self-reports from patients and/or their caregivers that are cumbersome to administer, largely subjective, and open to challenges in communication/interpretation, or (2) neuropsychological testing batteries that focus on basic cognitive outcomes that cannot delineate specific cognitive domains or mechanism. For example, a recent meta-analysis of 52 studies found that 1 in 3 breast cancer survivors may have significant cognitive impairment, although the prevalence was higher when subjective self-reported measures were considered (Whittaker et al., 2022). Traditional neuropsychological tests can only parse basic response data (i.e., accuracy, response time [RT]). Processing speed might be an outcome of cognitive impairment, as opposed to a discrete cognitive domain, since cognitive slowing can be caused by impairments in attention, executive function, and/or working memory. Furthermore, the correlation between subjective perceptions of cognitive decline in cancer patients and objective performance on neuropsychological testing is not supported. Rather, it seems the self-perception of problems with cognitive functioning in patients is correlated with psychosocial factors, such as depressed mood and anxiety (Boscher et al., 2020; Hutchinson et al., 2012; Whittaker et al., 2022). Thus, subjective reports might not be capturing true neurocognitive dysfunction, and neuropsychological testing provides limited information that potentially lacks sensitivity to capture deficits in specific neurocognitive domains. Collectively, this suggests the need to adopt new tools to examine neurocognitive functioning among non-CNS cancer survivor communities.
One method of cognitive neuroscience that is highly apt to objectively delineate specific cognitive domains and associated neurobiological dysfunction is neurophysiology, specifically Event-Related Potentials (ERPs; Luck, 2014). Cognitive ERPs provide gold-standard temporal resolution to the millisecond, in addition to source localization estimation of < 1 cm (Tivadar & Murray, 2019). Understanding the temporal mechanism of cognitive processing provides an efficient, non-invasive tool for studying the brain’s synaptic function, that orchestrates a multi-leveled system of neuronal assemblies (Badin et al., 2017). Technically, ERPs reflect the instantaneous summation of post-synaptic inhibitory and excitatory membrane potentials of pyramidal cells in the neocortex (Luck, 2014). Measuring the temporal trajectory of neurocognitive (dys)function is particularly advantageous since cognitive processing is extremely fast (and thus not fully capturable via other neuroimaging methods), where the temporal encoding of information is essential for synaptic plasticity during executive functioning, emotion control, inhibition, learning, and memory processing. ERP components in humans can be divided into two broad categories: (1) early “sensory” or “exogenous” components that peak amplitude before (or around) ~ 100 ms post-stimulus, and (2) late “cognitive” or “endogenous” components that peak from 100 ms onwards and represent more elaborative information processing (Thompson et al., 2020; Tivadar & Murray, 2019). Impairment or dysregulation of ERP components can represent phenotypic markers of neurocognitive deficit and synaptic dysfunction. As such, ERPs have been particularly useful in the assessment and diagnosis of Alzheimer’s Disease (Golob et al., 2009; Olichney et al., 2011), Mild Cognitive Impairment (MCI) (Correa et al., 2018; Olichney et al., 2005; Papaliagkas et al., 2008; Waninger et al., 2018), and Traumatic Brain Injury/Stroke (Cavanagh et al., 2019; Lowry et al., 2021; Miranda et al., 2020).
The majority of cognitive investigations in cancer survivorship to date are in lung and breast cohorts (Hervey-Jumper & Monje, 2021; Pendergrass et al., 2018). Here, we investigate survivors of head and neck cancers, a less studied non-CNS cancer survivor cohort, to contribute to the evidence base. Head and neck cancers are the 6th most common cancer worldwide (Vigneswaran & Williams, 2014), with approximately 890,000 new cases diagnosed each year, which is predicted to rise to 1.08 million by 2030 (Johnson et al., 2020). Head and neck cancers can include tumors of the larynx, pharynx, oral cavity paranasal sinus, and saliva gland, encompassing the oropharyngeal, nasopharyngeal, hypopharyngeal, laryngeal, lip and oral cavity, salivary gland, and paranasal sinus and nasal cavity, cancers. Developments in screening and treatments mean prognoses have improved, and fortunately, around 40% of patients with locally advanced head and neck cancer are disease free at 5 years (Cohen et al., 2016). This number is expected to also rise due to increased long-term survival in patients with human papillomavirus (HPV)-associated with oropharyngeal cancer. Mirroring other non-CNS cancer populations, significant impairments appear to have devastating late effect impacts on head and neck cancer survivors’ lives (Cocks et al., 2016; Murphy et al., 2019). A large portion of research and clinical care focuses on physical rehabilitation in the head and neck cancer survivor cohort, although data suggest that survivors commonly experience severe persistent cognitive impairments following remission (Bond et al., 2016; Gan et al., 2011; Tang et al., 2012; Yuen et al., 2008; Zer et al., 2018).
The extant evidence base examining neurocognitive deficits in head and neck cancer survivors is sparse. An early study reported unsafe driving simulator performances (i.e., average speed, mean brake reaction time, number of fatal collisions, steering variability) in head and neck cancer patients following cancer radiation treatment to the temporal lobes (Yuen et al., 2008). Increased risk of incidental irradiation injury to nerve fibers and vasculature in the temporal lobes during head and neck cancer treatment was also supported in a following study using neuropsychological testing batteries (Gan et al., 2011), that reported correlations between radiation dose to the temporal lobes and impaired memory, versus radiation dose to the cerebellum and impaired dexterity. A later study examined 46 recurrence-free nasopharyngeal carcinoma patients with suspected radiation injury exposed to radiation treatment, compared with 46 matched controls who were nasopharyngeal carcinoma patients post-radiation treatment without suspected radiation injury (Tang et al., 2012). While all patients (with and without radiation injury) presented psychological symptoms of anxiety and depression, patients with radiation injury also showed cognitive dysfunction, where the severity of cognitive impairment significantly correlated with the severity of abnormalities of white matter or temporal lobe radio-necrosis. To note, a review assessing data from more than 300 patients exposed to radiotherapy and/or chemotherapy proposed that a key difference between patients who develop radiation injury versus not may be in the use of concomitant platinum-based chemotherapy. Although platinum-based cytotoxic drugs have limited ability to cross the blood-brain barrier, conventional doses do reach concentrations within the brain to cause significant toxicity (Lange et al., 2019). More recent studies using neuropsychological testing have reported verbal-domain impairments in head and neck cancer patients receiving primary or adjuvant chemoradiation (Bond et al., 2016), and global neurocognitive impairment in patients with histologically proven head and neck cancer exposed to chemoradiotherapy or radiotherapy, compared with healthy controls (Zer et al., 2018). Interestingly, neurocognitive impairment was not associated with baseline cytokine markers, suggesting different methods are needed to explore this domain, and possibly provide predictive information.
Here, we leverage electrophysiological ERPs toward identifying candidate phenotypic markers of neurocognitive impairment in head and neck cancer survivors with potential scope for the greater oncology field, that manifest as the devastating effects on everyday cognitive functions involving attention, sensory encoding, executive functioning, multi-tasking, memory, and comprehension. Specifically, we used standardized Go/Nogo testing paradigms concomitant to EEG recording to investigate early exogenous ERP components gauging sensory perception and early attention gating (C1, P1, N1); the germane ERP component related to executive function and inhibitory control (N2); and later ERP component associated with conscious attention processing and conscious evaluation (P3). We also examined the N4 ERP component based on previous reports that cancer survivors displayed deficits in language processing (Bond et al., 2016; Gan et al., 2011; Tang et al., 2012; Yuen et al., 2008; Zer et al., 2018). We hypothesized that head and neck cancer survivors would yield differentiated amplitudes in the aforementioned ERP components compared with age/gender/education-matched healthy controls and that differences in ERP morphology would also correlate with their clinical symptomatology.
Methods
Design and Sample
This was a retrospective cohort study of head and neck cancer survivor patients and matched healthy controls. Eligibility criteria for head and neck cancer survivors were as follows: (1) age ≥ 21 years; (2) head and neck cancer of the larynx, pharynx, oral cavity paranasal sinus, salivary gland, or unknown primary; (3) any histology of any epithelial origin; (4) completed therapy a minimum of 6 months prior to study entry; (5) at least two systemic symptoms on the general symptom subscale of the Vanderbilt Head and Neck Symptom Survey (VHNSS-GSS; Cooperstein et al., 2012); (6) no history of neurodegenerative disease, unrelated to cancer history/treatment; (7) English language ability to understand instructions and be able to provide informed consent. Exclusion criteria for all participants were: (1) alcohol/substance abuse/dependence within the last 6 months; (2) current or previous co-morbid bipolar disorder, psychosis, obsessive-compulsive disorder, eating disorder, personality disorder; (3) neurological disorders unrelated to cancer and its treatment (e.g., ADHD, ASDs, epilepsy); (4) learning difficulties impeding comprehension for providing consent and/or during experimentation. Head and neck cancer survivor patients were recruited from the Vanderbilt Department of Otolaryngology, Vanderbilt Head and Neck Cancer clinics. Healthy controls were recruited via the VUMC recruitment listserv and ResearchMatch. All research procedures were in accordance with the Vanderbilt University Institutional Review Board and VICC’s Scientific Review Committee approval.
Procedure
Potential patients were initially identified by a clinical oncologist and followed up by the research team for specific research-based screening. Potential healthy control participants were provided a link to HIPAA-compliant RedCap screening surveys to ascertain eligibility criteria. Informed written consent was obtained from each participant prior to experimental testing, which included completing self-report surveys during EEG cap set-up, followed by completing neurocognitive experimental paradigms concomitant to EEG recording. The experiment sequence order was randomized across all participants. Participants were instructed to keep muscular activity relaxed (e.g., forehead, shoulders), and eye movements minimal, as much as possible during EEG recording periods. At the end of the study, all participants received financial compensation as a token of gratitude for their time.
Experimental Paradigms
EEG recordings began with a 6-min baseline, comprising a 3-min eyes-open followed by a 3-min eyes-closed standardized protocol. Participants then underwent cognitive and emotional variants of the Go/Nogo neurocognitive paradigms that are well-validated to assess sustained attention, executive/inhibitory control, and evaluative processing, as well as elicit adequate trials to ensure averaged ERP measures represent ecological reliability and validity. Specifically, the cognitive Go/Nogo paradigm has been validated in patients with attention deficits by our team (Schoenberg et al., 2014), and the emotional Go/Nogo paradigm in patients with mood disorders by our team, using EEG methods (Schoenberg, 2014; Schoenberg & Speckens, 2014). For the purpose of this study, we adapted the emotional Go/Nogo paradigm to include chronic pain-related valenced words balanced with positive and neutral valenced words. The word list protocol was matched for usage and length and tested previously in a pilot sample of chronic pain patients.
The cognitive Go/Nogo paradigm comprised the sequential presentation of five letters (A, F, H, Y, X): h = 2 cm, w = 1.5 cm, white on black background. Patients were instructed to press a button on a response pad as quickly and accurately as possible whenever they saw the letters “A,” “F,” “H,” or “Y” (396 Go stimuli), and do not press whenever they saw an “X” (99 Nogo stimuli). Overall, 495 stimuli (20% inhibition rate) were presented in 3 × 165 stimuli blocks, with rest intervals between each block. Stimulus duration = 500 ms, random interstimulus interval [ISI] = 750 – 2,200 ms.
The emotional Go/Nogo paradigm was adapted from our previous study in a depressed population (Schoenberg & Speckens, 2014, 2015). The present task involved the sequential presentation of positive, neutral, and pain-related words. Pain-related words were used in this study, replacing the depressogenic words used in our mood disorder study, because the experience of pain is a significant part of chronic systemic symptomatology in non-CNS cancer survivors and was potentially more salient in terms of valence elicitation. Standardized word stimuli were sourced from; 78 pain-related words from the McGill Pain Questionnaire (Melzack, 1987), 12 were derived from a previous pain study (Jensen et al., 2013), and 105 from an online survey created by our lab asking a diverse pool of participants including sufferers/caregivers/clinicians of chronic pain, to name words that they associated with pain. Positive and neutral words (200 words per emotional valence) were translated from standardized Dutch word databases into English from our previous study (Schoenberg & Speckens, 2014, 2015). The emotional Go/Nogo paradigm comprised 12 × 100 stimuli blocks with rest intervals between each block. Each block consisted of two possible Response types (Go or Nogo) [80 × Go − 20 × Nogo (20% inhibition rate)], constituting six possible “Go-Nogo” Condition Types: 1. Positive-Pain; 2. Positive-Neutral; 3. Pain-Positive; 4. Pain-Neutral; 5. Neutral-Positive; 6. Neutral-Pain. Each block type was repeated constituting 12 total randomly presented blocks. Stimulus duration was randomly presented (500–1,500 ms), with a randomized ISI (800–1,750 ms). Prior to each block standardized instructions were presented onscreen, verified verbally by the experimenter, indicating which word valence to press vs. not press in each block. Instruction presentation also provided rest periods.
Clinical Outcomes
The following self-report surveys were collected from all participants:
Vanderbilt Head and Neck Symptom Survey (VHNSS) version 2.0 plus General Symptom Survey (GSS) is a validated tool to measure symptom burden and functional deficits in head/neck cancer and its treatment (Cooperstein et al., 2012).
Neurotoxicity Rating Scale (NRS) is a self-report 37-item tool examining neurocognitive symptoms associated with neurotoxicity of medical treatment (Aldenkamp et al., 1995).
PROMIS-29 assesses seven domains (depression, anxiety, physical function, pain interference, fatigue, sleep disturbance, and participation in social roles/activities) using a 5-point Likert scale, across 29 items. It demonstrates excellent internal consistency and the ability to compare across conditions and with normalized data (Katz et al., 2017).
Behavior Rating Inventory of Executive Function – Adult version (BRIEF-A) is a standardized 75-item measure that captures views of an adult’s executive functions or self-regulation in his or her everyday environment. It measures nine non-overlapping theoretically and empirically derived clinical domains: Inhibit, Self-Monitor, Plan/Organize, Shift, Initiate, Task Monitor, Emotional Control, Working Memory, and Organization of Materials (Roth et al., 2005).
EEG Recording and Acquisition
EEG data were acquired at the Osher Center for Integrative Medicine at Vanderbilt neurophysiology laboratory using the research-grade BrainAmp DC hardware and Brain Vision Recorder 1.24 software (http://BrainProducts.com). Electrode set-up comprised 64 electrode montages according to the international 10–20 localization system. AFz electrode was used as the ground electrode, and the FCz electrode as the recording reference electrode. The vertical and horizontal ocular activity was calculated by bipolar derivations of electrooculogram (EOG) signals recorded above and below the right eye (vEOG) and 1 cm to the outer canthi of each eye (hEOG). vEOG and hEOG were used for ocular correction. Impedance level was maintained ≤ 10 KΩ. The electrical signal was continually sampled at a digitization rate of 500 Hz, band-pass filtered between 0.1 and 100 Hz.
EEG Signal Processing
ERP analysis was conducted using Brain Vision Analyzer 2.2.2. Data were referenced to the average of all scalp electrodes producing an average reference, including the reconstruction of the signal of the recording reference electrode (FCz) with the new average reference so that the channel could be re-introduced into the signal montage. Data were filtered between 0.1 and 30 Hz (24 dB/octave slope), using zero-phase shift bandpass IIR Butterworth filters, and a 60 Hz Notch filter. Ocular artifacts were corrected using a widely standardized regression method for ERP data using vEOG and hEOG derivations (Gratton & Coles, 1983). The signal was visually inspected for muscular and other artifacts, and epochs were rejected if data exceeded 100 μV. Data were segmented into stimuli-locked All-Go (and also by valence for the emotional Go/Nogo), and All Nogo trials; and response-locked to Correct Hits to Go, and False Alarms to Nogo. Segments were baseline corrected −200 ms to 0 ms, prior to averaging, and grand averaging for each component. Adaptive mean values (3−/+ data points/12 ms) around the identified peak amplitudes and latencies, at maximal electrode sites within the sample for the C1 (50–100 ms negative polarity), P1 (80–150 ms positive polarity), N1 (80–200 ms negative polarity), N2 (200–350 ms negative polarity), P3 (250–550 ms positive polarity), N4 (350–550 ms negative polarity), ERP components were extracted. Electrode sites that were maximal for each component in our sample were in line with other studies’ maximal topographies. Furthermore, the adaptive mean amplitude is optimal since it offsets low signal-to-noise ratio more reliably while maintaining individual ERP variability (Clayson et al., 2013).
Behavioral Data
Group means and standard deviations for the reaction times of cancer survivors and healthy controls were calculated. Trials with a response < 150 ms or > 2 standard deviations above the group mean were removed. For each participant, the total number of Go trials and a total number of Nogo trials were computed. Correct Go responses (i.e., correctly pressing the button during the presentation to Go stimuli) and false alarms (i.e., pressing the button incorrectly for Nogo stimuli) were recorded. Hit rates, defined as the proportion of correct responses during Go trials, were calculated by dividing the total number of correct responses during Go trials by the total number of Go trials. The false alarm rate, defined as the proportion of responses during Nogo trials, was calculated by dividing the total number of false alarms by the total number of Nogo trials. The coefficient of variability (CV), used to measure the variation in responses during Go trials, was calculated by dividing the average reaction time across all correct Go trials by the standard deviation of correct Go trials. Higher CV indicated more variability in RTs. D-prime (d′), a sensitivity index used to measure performance on a Go/Nogo task while accounting for bias, was calculated by subtracting the z-standardized false alarm rate from the z-standardized hit rate.
Statistical Analysis
Our primary hypothesis was to investigate ERP component differences between non-CNS cancer survivors versus matched healthy controls. Repeated-measure analysis of variance (ANOVA) with Bonferroni correction were conducted separately for each experimental design (cognitive or emotional Go/Nogo), and amplitude (μV) or latency (ms) data. Due to the discrete functional significance of the ERP components investigated (C1, P1, N1, N2, P3, N4), they were not run through the same statistical model, but analyzed separately. Specifically, repeated-measure ANOVA with two Electrode (FCz, Cz) × two Condition (Correct Hits to Go, All Nogo) within-subjects factors, and two Group (non-CNS cancer survivors, matched healthy controls) as a between-subjects factor, were conducted on the C1 ERP during cognitive Go/Nogo paradigm. The same model was applied for the other components, except the within-subjects factor of Electrode was: POz, Oz for P1; Fz, POz, Oz for N1; FCz, CPz for N2; Pz, POz, Oz for P3; Cz, CPz for N4. For the emotional Go/Nogo paradigm, an additional within-subjects factor of three Valence (positive, pain, neutral) was included. Where the assumption of sphericity was violated, Greenhouse Geisser correction was applied. Effect sizes (η2) are reported as useful information for future studies. Significant main effects/interactions were unpacked further using Independent-Samples t-tests. Where ERP component data were significant, the interaction between ERP values and self-report executive function assessment; and ERPs with the clinical toxicity and lived experiences of physical, mental, and social health; were examined using point-biserial correlation, split by Group, to cross-correlate categorical with continuous data. Because correlation coefficients are effect sizes, a correction was not applied to p values to account for multiple comparisons, rather the adjusted R squared (R2adj) was calculated and also reported. Behavioral data were analyzed using repeated-measure ANOVA with Bonferroni correction in a two Condition (hit rates to Go, false alarm rates to Nogo), three-Valence (neutral, positive, pain) within-subjects factors, and two Group (non-CNS cancer survivors, matched healthy controls) as a between-subjects factor for the emotional Go/Nogo paradigm, where this model was also conducted on RT data in a separate analysis. Group differences for the remaining behavioral variables of D-prime and CV, and all measures for the cognitive Go/Nogo paradigm, were analyzed using Independent-Samples t-tests. One-way ANOVA compared differences in age and education between the two groups (cancer survivors vs. matched healthy controls), and to check no significant differences in categorical demographic data (such as gender and race) between groups, the chi-square (χ2) co-efficient was used. Clinical survey data was analyzed using independent samples t-tests.
Results
A retrospective cohort study design compared twenty-one participants; 11 head and neck cancer patients who were cancer free and had a mean of 4.6 years (range 1–16) from their final successful treatment versus 10 healthy control participants. Table 1 presents demographics and Table 2 clinical symptom measures of the study sample. Behavioral performance data for the cognitive and emotional Go/Nogo paradigms are reported in Tables 3 and 4, respectively. Table 5 provides an overview of all the main ERP findings in cancer survivors compared to matched healthy controls. Tables 6 and 7 provide the number of trials averaged for each ERP for the cognitive and emotional Go/Nogo paradigms, respectively.
Cognitive Go/Nogo Paradigm: ERP Amplitude Data
Main effect of Condition, F(1, 18) = 5.426, p = .032, η2 = .232, Condition × Group interaction, F(1, 18) = 6.343, p = .021, η2 = .261, and effect of Electrode, F(1, 18) = 8.864, p < .001, η2 = .330, were observed for N1 ERP amplitudes. Post hoc t-tests revealed cancer survivors yielded higher N1 amplitudes (more negative potential) compared to healthy controls for Correct Hits to Go trials at the frontal electrode (Fz), t(18) = −2.395, p = .028 (−2.22 μV vs. −.82 μV).
Main effects of Condition, F(1, 18) = 459.209, p < .001, η2 = .962, Electrode × Group interaction, F(1, 18) = 439.116, p < .001, η2 = .961, and effect of Electrode, F(1, 018) = 456.730, p < .001, η2 = .962, were observed for N2 ERP amplitudes. Post hoc t-tests revealed cancer survivors yielded lower amplitudes for Nogo trials at the central-parietal (CPz) electrode, t(19) = 2.370, p = .029 (−1.32 μV vs. −3.70 μV).
No significant main effects/interactions were observed for C1, P1, P3, and N4, amplitudes.
Cognitive Go/Nogo Paradigm: ERP Latency Data
Significant findings were exclusive to the P3 ERP component. The main effects of Group, F(1, 18) = 8.655, p = .009, η2 = .325, and Electrode F(2, 36) = 9.600, p < .001, η2 = .348, revealed that compared with healthy controls, cancer survivors yielded significantly shorter (faster) latencies at; parietal electrode (Pz), t(19) = −2.078, p = .052 (346.55 ms vs. 428.80 ms), for All NoGo trials, and occipital electrode (Oz), t(18) = −2.350, p = .030 (266.00 ms vs. 341.40 ms) for Correct Hits to Go. At the posterior-occipital electrode (POz), cancer survivors also yielded shorter (faster) latencies for Correct Hits to Go, t(18) = −3.503, p = .003 (271.60 ms vs. 383.20 ms), and All NoGo trials, t(19) = −2.138, p = .046 (302.91 ms vs. 373.00 ms).
No significant main effects/interactions were observed for C1, P1, N1, N2, and N4, latencies.
Emotional Go/Nogo Paradigm: ERP Amplitude Data
An Electrode × Valence × Condition × Group interaction, F(2, 38) = 3.494, p = .040, η2 = .155, was observed for the C1 ERP component. Follow-up tests revealed that cancer survivors yielded lower amplitudes compared to healthy controls for the All Positive-Nogo condition at frontal (Fz) electrode, t(19) = −3.050, p = .007 (−1.25 μV vs. −0.31 μV).
A Condition × Valence × Group interaction F(2, 38) = 3.282, p = .048, η2 = .147, and main effect of Electrode F(2, 38) = 4.016, p = .026, η2 = .174, were observed for the N1 ERP component. Post hoc tests revealed that cancer survivors yielded lower amplitudes at the posterior occipital electrode (POz) for All PAIN-NoGo trials, t(19) = 2.174, p = .042 (−0.37 μV vs. −2.73 μV); in addition for Correct Hits to POSITIVE-Go trials, t(19) = 2.258, p = .036 (−0.56 μV vs. −2.35 μV). Interestingly, despite being only trend significance (p = .066), Correct Hits to PAIN-Go showed an opposite pattern of higher amplitudes in cancer survivors compared to controls (−2.38 μV vs. −1.25 μV).
No main effects/interactions of significance were observed for P1, N2, P3 (aside from Valence effect only), and N4 amplitudes.
Emotional Go/Nogo Paradigm: ERP Latency Data
An Electrode × Valence × Group interaction, F(2, 38) = 3.322, p = .047, η2 = .149, was observed for the N1 ERP component, which showed significant post hoc findings for PAIN-related trials. At the parieto-occipital (POz) electrode cancer survivors yielded faster ERP latencies for All PAIN-Nogo trials, t(19) = −3.117, p = .006 (104.36 ms vs. 147.80 ms). Despite only being trend (p = .057), the converse was true for Correct Hits to PAIN-Go stimuli at the frontal (Fz) electrode (150.36 ms vs. 120.60 ms).
A main effect of Group F(1, 19) = 8.019, p = .011, η2 = .297 for the N4 ERP component, highlighted that cancer survivors yielded significantly longer (slower) latencies at the central electrode (Cz) for Correct Hits to POSITIVE-Go, t(19) = 2.532, p = .020 (451.27 ms vs. 399.20 ms), Correct Hits to PAIN-Go, t(19) = 4.251, p < .001 (477.64 ms vs. 385.80 ms), and Correct Hits to NEUTRAL-Go stimuli, t(19) = 2.917, p = .009 (464.36 ms vs. 394.00 ms). Furthermore, cancer survivors yielded significantly longer (slower) latencies for Correct Hits to PAIN-Go stimuli t(19) = 2.841, p = .010 (463.45 ms vs. 393.40 ms) at the central-parietal (CPz) electrode.
No main effects/interactions of significance were observed for C1, P1, N2, or P3 (aside from Electrode effect only), latencies.
Correlations for Cognitive Go/Nogo: ERP × Clinical Data
Several significant correlations pertained to the P3 ERP component latencies and clinical/functional assessments in cancer survivors. Specifically at Oz for All Nogo trials and the PROMIS-Depression scores, r = −.611, p = .046, R2adj = 41.37; PROMIS-Social Functioning scores, r = −.723, p = .012, R2adj = 57.97; and Behavior Rating Inventory of Executive Function (BRIEF) “Emotional Control” domain, r = −.676, p = .022, R2adj = 50.67. These correlations were not significant in the healthy control group. Furthermore, at POz for All Nogo trials and the PROMIS-Sleep Disturbance scores, r = .609, p = .047, R2adj = 41.1; PROMIS-Depression scores, r = −.644, p = .032, R2adj = 45.96; and PROMIS-Social Functioning scores, r = −.640, p = .034, R2adj = 45.4. Conversely, correlation in the opposite direction was observed in healthy controls for POz for All Nogo trials and PROMIS-Sleep Disturbance scores, r = −.703, p = .023, R2adj = 54.79. Finally, N1 ERP component amplitudes at Fz for Correct Hits to Go correlated with the Behavior Rating Inventory of Executive Function (BRIEF) “Shift” domain, r = −.745, p = .013, R2adj = 61.55, in cancer survivors, that was not observed in healthy controls.
Correlations for Emotional Go/Nogo: ERP × Clinical Data
Significant correlations were exclusive to the N1 ERP component amplitudes at POz for Correct Hits to POSITIVE-Go with various domains on the Behavior Rating Inventory of Executive Function (BRIEF) in cancer survivors: (a) “Inhibit”, r = .701, p = .016, R2adj = 54.48; (b) “Initiate”, r = .704, p = .016, R2adj = 54.95; and (c) “Organization of Materials”, r = .648, p = .031, R2adj = 46.54. N1 amplitude at POz for Correct Hits to POSITIVE-Go also correlated with PROMIS Anxiety scores, r = .845, p = .001, R2adj = 79.21, in cancer survivors. These correlations were not apparent in the healthy control group.
Discussion
We set out to examine whether; (1) non-CNS cancer survivors > 6 months from post-treatment show neurocognitive impairments compared with age/gender/education-matched healthy controls (to note in this cohort, cancer survivors were a mean 4.6 years out from treatment completion), (2) such deficits can be captured and delineated by summated post-synaptic activity of local pyramidal neural populations measured using electrophysiological ERPs, and (3) neurocognitive impairments pertain to specific functioning domains.
Very Early Attention and Sensory Perception Components (C1, N1, P1 ERPs)
Non-CNS cancer survivors generally showed significant differences, in both amplitude and latency, across the very early ERP components, that is, exogenous potentials that capture neuronal responses to sensory structure reflecting a function of processing to the physical properties of external stimuli (Sur & Sinha, 2009). Any deficit in very early components can thus have knock-on effects down the neural processing chain. The C1 ERP is the first in this series of exogenous components related to visual attention, followed by the P1-N1 ERP complex (C1-P1-N1) (Ahmadi et al., 2018). The neural generator of the C1 ERP has been largely sourced to the primary visual cortex (V1) (Di Russo et al., 2002), associated with both cognitive load (perceptual/attentional) and the type of attention (voluntary/involuntary) (Slotnick, 2018) involved in the experimental task eliciting C1. Higher perceptual load within the experimental design (i.e., targets in a smaller region of the visual field) means greater attentional allocation to targets, whereas higher attentional load (i.e., targets in the central visual field) may lead to lesser attentional allocation to distractors in other visual locations (Slotnick, 2018). Here, we used central target (and non-target) locations of the visual field for stimuli presentation, although we were not specifically tapping spatial changes in visual processing. Lower C1 amplitudes were exclusive to positive stimuli during the emotional Go/Nogo in non-CNS cancer survivors. This suggests that non-CNS cancer survivors displayed decreased “pre-attentive” processing within the visual field, which interestingly appeared specific to positive information. The latter points to a non-specific slowing in the sensory gating mechanism of encoding physical properties of external stimuli.
Non-CNS cancer survivors yielded a dispersed pattern for the N1 ERP component. With the cognitive Go/Nogo, amplitudes were higher on correct responses to Go stimuli, which also correlated with the “Shift” subdomain of the Behavior Rating Inventory of Executive Function (BRIEF) that assesses the ability to (1) switch/alternate attention, (2) make transitions, (3) tolerate change, (4) problem-solve flexibly, and (5) change focus from one mindset/topic to another (Roth et al., 2005). However, N1 amplitudes were modulated by valence and electrode site in the emotional Go/Nogo, and specifically correlated with the BRIEF domains of “Inhibit” (behavioral regulation, such as the ability to inhibit, resist, or not act on one’s impulses, appropriately stop own behavior at the proper time), “Initiate” (the ability to begin tasks/activities, and independently generate ideas), and “Organization of Materials” (organization in an adult’s everyday environment with respect to the orderliness of living, work, and other spaces, keeping track of important personal belongings) (Roth et al., 2005), and scores on anxiety. Neurophysiologically, N1 amplitudes were lower at posterior electrode sites for pain Nogo and positive Go stimuli yet showed the inverse of higher amplitudes for correct hits to pain Go stimuli. Furthermore, these amplitude patterns were mirrored in N1 latency data in head and neck cancer survivors, which were also modulated by valence. For example, N1 peaked significantly faster to pain-related stimuli, and significantly slower to positive-related stimuli in non-CNS cancer survivors. Both the P1 and N1 ERPs are shown to be generated in extra-striate cortical regions and are modulated by top-down control of attention (Hsu et al., 2014). The N1 serves in an early gating system that “matches” presently attended stimuli with a template from previously processed stimuli, facilitating further perceptual classification and discrimination processing (Ernst et al., 2013; Vogel & Luck, 2000). These findings suggest that non-CNS cancer survivors required greater neuronal allocation from this system to the “Nogo” conditions yet still made errors (i.e., did not inhibit their response), as demonstrated by higher amplitudes to errors made on positive and neutral trials. Whether this was due to an impairment in pre-attentive stages of processing or subsequent inhibitory control, may be delineated by examining the later stage components, Nogo N2 and P3, as follows.
Inhibitory Control and Conflict Monitoring (N2 ERP)
Reduced amplitudes and longer latencies, independent of emotion processing, characterized the N2 morphology of non-CNS cancer survivors. The Nogo N2-P3 complex reflects an inhibitory gating circuit where the Nogo-N2 represents the early regulatory stage of conflict monitoring, compared with the “closure” Nogo-P3 potential of this system that is associated with cognitive and motor control to inhibit response (Schoenberg et al., 2014). The later Nogo-P3 component showed no amplitude (functional) difference between survivors and controls, suggesting an impairment of the earlier regulatory gating component of this inhibitory system. This is supported by reduced Nogo-N2 amplitudes in survivors since increased amplitudes are associated with enhanced conflict monitoring and response inhibition capacity (Schoenberg et al., 2014). Reduced amplitudes have also been found in various conditions affecting cognition, including attention deficit disorders, Alzheimer’s disease, stroke, and impaired cognitive control in addictions (Buzzell et al., 2014; Olichney et al., 2011; Schoenberg et al., 2014).
Conscious Attention and Evaluation (P3 ERP)
Unlike early (C1-P1-N1) and middle (N2) latency ERPs, no differences in amplitude were apparent between non-CNS cancer survivors and matched healthy controls for later stages involving evaluation and semantic processing. The P3 is an endogenous component that is not modulated by the physical properties of the external stimuli but rather is modulated by attention and later cognitive processing involved in the experimental paradigm investigated (Ahmadi et al., 2018). P3 latency generally increases with the complexity of the stimuli evaluation and decision processes demanded by the task (Hsu et al., 2014). Later components such as N2 and P3 appear to show enhanced sensitivity to dementia than early sensory components (Hsu et al., 2014). This pattern was not reflected in our findings, suggesting that any neurocognitive impairment in non-CNS cancer survivors following their cancer/treatments does not follow a clear neurodegenerative trajectory akin to dementia, albeit more likely lesion-based damage.
Furthermore, our P3 latency findings did not align with other cancer cohort studies, especially examining the effects of chemotherapy and/or radiotherapy. A previous study in another non-CNS cancer type, breast cancer, reported prolonged P3 latencies that were of higher likelihood in younger (< 50 years) than older patients (Xu et al., 2020), particularly during treatment and up to two years follow-up. The authors concluded risks of chemotherapy upon cognitive functioning decreased with age (≥ 60 years), where younger patients with breast cancer receiving chemotherapy might display neurocognitive impairments for many years following remission. An ERP study examining cognitive function in childhood leukemia survivors (mean age at testing 15.5 years), also reported slower P3 latencies compared with healthy controls, that were not observed in earlier components such as the mismatch-negativity (MMN; e.g., Lähteenmäki et al., 2001). Prolonged P3 latencies are associated with neurocognitive impairments in frontal lobe lesions, incurred from stroke, brain tumors, or traumatic brain injury (Lai et al., 2013). In our non-CNS cancer survivor sample, no significant P3 differences between survivors and healthy controls were evident in the emotional Go/Nogo and were faster – not slower – at posterior electrodes during the cognitive Go/Nogo. Moreover, these P3 latency differences in cancer survivors correlated with depression, sleep health, and social functioning scores, in addition to the behavioral executive functioning domain of “Emotional Control” (executive functioning with regards to emotion processing, the ability to modulate one’s emotional responses) on the BRIEF (Roth et al., 2005).
Our window for maximal P3 peak was within 250–550 ms. While the P3 is considered a “300 ms” component, variability is common in clinical cohorts. Looking at the waveform morphologies of the cancer survivors in our sample, the early subcomponents of the waveforms had a slightly later morphology, compared with earlier morphology compared to controls as they moved into the 200 ms onward subcomponents, that is, see patterns in Figures 1 and 2 for both Go and Nogo trials. Thus, it does not seem probable that the faster latencies of the P3 ERP are residue from a prolonged P2, for example. It is also curious that when we look at the performance data, the cancer survivors showed significantly slower reaction times across the board compared with healthy controls, suggesting a discrepancy between neurophysiological processing and overt behavioral response. The P3 is a complex ERP with multiple functional representations depending upon experimental properties and context. One aspect to consider is that P3 latency represents the time required by the participant to evaluate the stimulus independent of the time and systems required to process a response (Duncan-Johnson & Kopell, 1981). Overall, faster P3 latencies are accompanied by larger P3 amplitudes to be considered a marker of efficient information processing. No amplitude differences were apparent between the groups. Moreover, when we look at the performance data, cancer survivors yielded fewer false alarms (incorrect responses to Nogo stimuli), yet also yielded fewer correct hits to Go stimuli, which was markedly apparent in the emotional Nogo. In sum, these findings suggest cancer survivors were not processing information more efficiently, albeit more cautiously from a behavioral perspective. Neurophysiologically, the P3 ERP is aligned with context-updating and content-closure dimensions of cognitive processing (Dinteren et al., 2014). In sum, this updating mechanism was not sustained in cancer survivors such that perception and evaluation processing of the stimuli was only possible when highly match to an already concretely established mental representation of the stimuli, essentially reflecting very low cognitive flexibility and ability to sustain attention.
Semantic Language Processing (N4 ERP)
As to be expected, there were no significant findings related to the cognitive (letter stimuli) Go/Nogo paradigm. However, the emotional (word stimuli) Go/Nogo paradigm yielded a pattern of overall longer latencies for all stimulus types (pain, positive, and neutral). This would suggest a global slowing in semantic language processing in non-CNS cancer survivors. A large portion of the previous research examining neurocognitive deficits specifically in head and neck cancer has focused on language processing and reading/verbal-based neuropsychological testing (Bond et al., 2016; Gan et al., 2011; Tang et al., 2012; Yuen et al., 2008; Zer et al., 2018). However, our data shed further light on these findings since specifically N4 amplitudes were not affected in non-CNS cancer survivors, suggesting no specific language-based impairment, that was the conclusion from many of these earlier studies using neuropsychological testing batteries. The emotional Go/Nogo using valenced word stimuli requires semantic linguistic processing, which would be reflected in the N4 (and evaluative P3) ERP, which did not yield significant findings. Rather impairments to early visual attention, conflict monitoring, and inhibitory control, as found in the early/middle ERP components in our study, may perhaps better explain any cognitive slowing shown in semantic language-based processing later down the processing chain. This also highlights the specialized advantage of using cognitive neurophysiology to delineate potential neurocognitive damage in non-CNS cancer survivors.
Limitations
To our knowledge, this is the first study to leverage multiple electrophysiological Event-Related Potentials/ERPs components towards the identification and delineation of impairment in specific neurocognitive domains in head and neck cancer survivors, a non-CNS cancer type. Although ERPs are a particularly sensitive neuroimaging technique, our presented findings should be assessed conservatively due to the moderately small sample size that may have contributed to the relatively small to mid-range effect sizes. This report provides useful data towards further replication, hypothesis generation, assessment methods testing, and treatment innovation for cognitive impairment rehabilitation within the non-CNS cancer community. Participants were disproportionately male, in line with head and neck cancer incidence rates in North America, and an older age cohort. However, data were compared with a age/gender/education-matched healthy control group, which although had slightly more years of education than survivors, was not significant. This provided an opportunity to rule out age or education-related differences in cognitive function between groups, supporting the neurocognitive impairments in survivors were a result of cancer and/or its treatment.
Summary and Future Directions
Cancer diagnosis and treatment are life-altering, and survivorship remains highly challenging despite improved cancer survival outcomes. This first examination of discrete cognitive ERPs in a cohort of non-CNS cancer survivors (previously afflicted with head and neck cancer) shows significant neurocognitive impairments in executive functioning related to sensory gating of attention, perceptual encoding, conflict monitoring, and later onset inhibitory control. Impairments in these cognitive domains will detrimentally impact daily living. Executive functioning is known to decline as part of healthy cognitive aging (Kirova et al., 2015), although we did not see the same ERP morphologies in our age-matched healthy control group. Our findings are particularly interesting because aberrations in amplitude morphology were not necessarily global, rather in some instances were emotion-related, specifically to the processing of positive and pain stimuli. In line with this, non-CNS cancer survivors displayed slower reaction times behaviorally across the board, and longer latencies for many ERP components, yet this pattern was the contrary for pain-related stimuli during very early processing ERP components (such as the N1), with survivors yielding considerably shorter latencies to pain trials and significantly longer latencies to positive trials. These findings were not reflected in quicker reaction times in survivors, however, suggesting the suitability of neurophysiological ERPs in gauging neurocognitive differences.
Amplitude morphology of ERPs indexing later evaluative functions remained intact, suggesting that any latency slowing in these ERPs may be a knock-on effect from deficits in earlier components of the executive function processing chain, particularly related to visual attention and gating systems. Furthermore, the ERP aberrations highlighted here did not follow a similar pattern of Alzheimer’s Disease, suggesting that any damage caused by cancer/its treatments mimics lesion damage as with stroke/epilepsy, as opposed to the onset of neurodegenerative decline. Further research leveraging cognitive neurophysiology is warranted. Albeit the initial evidence reported here suggests that neurocognitive rehabilitation programs targeting ERP component modulation, and/or underlying neurotransmission, have promise for specific impairments in the executive function of the non-CNS cancer survivor community.
Much gratitude and appreciation to all the participants of this research. We thank Natalie Lockney for proofreading the finalized manuscript.
References
(2018). Perceptual learning induces changes in early and late visual evoked potentials. Vision Research, 152, 101–109. https://doi.org/10.1016/j.visres.2017.08.008
(1995). The neurotoxicity scale: The validity of a patient-based scale, assessing neurotoxicity. Epilepsy Research, 20(3), 229–239. https://doi.org/10.1016/0920-1211(94)00082-8
(2017). Changes in cognitive functions and cerebral grey matter and their associations with inflammatory markers, endocrine markers, and APOE genotypes in testicular cancer patients undergoing treatment. Brain Imaging and Behavior, 11(3), 769–783.
(2017). The features and functions of neuronal assemblies: Possible dependency on mechanisms beyond synaptic transmission. Frontiers in Neural Circuits, 10, Article
114 .(2016). Neurocognitive function in patients with head and neck cancer undergoing primary or adjuvant chemoradiation treatment. Supportive Care in Cancer, 24, 4433–4442.
(2020). Perceived cognitive impairment in breast cancer survivors and its relationships with psychological factors. Cancers, 12(10), Article
3000 . https://doi.org/10.3390/cancers12103000(2014). The N2 ERP component as an index of impaired cognitive control in smokers. Neuroscience Letters, 563, 61–65. https://doi.org/10.1016/j.neulet.2014.01.030
(2019). ERPs predict symptomatic distress and recovery in sub-acute mild traumatic brain injury. Neuropsychologia, 132, Article
107125 . https://doi.org/10.1016/j.neuropsychologia.2019.107125(2020). Systematic review of cognitive sequelae of non-central nervous system cancer and cancer therapy. Journal of Cancer Survivorship, 14, 464–482.
(2013). How does noise affect amplitude and latency measurement of event-related potentials (ERPs)? A methodological critique and simulation study. Psychophysiology, 50, 174–186.
(2016). Palliative and supportive care in head and neck cancer: United Kingdom National Multidisciplinary Guidelines. Journal of Laryngology and Otology, 130(s2), S198–S207.
(2016). American Cancer Society head and neck cancer survivorship care guideline. CA: A Cancer Journal for Clinicians, 66(3), 203–239. https://doi.org/10.3322/caac.21343
(2012). Vanderbilt head and neck symptom survey version 2.0: Report of the development and initial testing of a subscale for assessment of oral health. Head & Neck, 34(6), 797–804.
(2018). Increased amplitude of the P3a ERP component as a neurocognitive marker for differentiating amnestic subtypes of mild cognitive impairment. Frontiers in Aging Neuroscience, 10, Article
19 .(2014). P300 development across the lifespan: A systematic review and meta-analysis. PLoS One, 9(2), Article
e87347 .(2002). Cortical sources of the early components in the visual evoked potential. Human Brain Mapping, 15(2), 95–111.
(1981). The Stroop effect: Brain potentials localize the source of interference. Science, 214, 938–940.
(2013). N1 and N2 ERPs reflect the regulation of automatic approach tendencies to positive stimuli. Neuroscience Research, 75, 239–249.
(2011). Cognitive functioning after radiotherapy or chemoradiotherapy for head-and-neck cancer. International Journal of Radiation Oncology and Biological Physics, 81(1), 126–134. https://doi.org/10.1016/j.ijrobp.2010.05.004
(2009). Cortical event-related potentials in preclinical familial Alzheimer disease. Neurology, 73, 1649–1655.
(1983). A new method for off-line removal of ocular artifact. Electroencephalography and Clinical Neurophysiology, 55, 468–484.
(2021). Unravelling the mechanisms of cancer-related cognitive dysfunction in non-central nervous system cancer. JAMA Oncology, 7(9), 1311–1312.
(2014). Both attention and precition are necessary for adaptive neuronal tuning in sensory processing. Frontiers Human Neuroscience, 8, Article
152 .(2012). Objective and subjective cognitive impairment following chemotherapy for cancer: A systematic review. Cancer Treatment, 38, 926–934.
(2013). The words patients use to describe chronic pain: Implications for measuring pain quality. Pain, 154(12), 2722–2728.
(2020). Head and neck squamous cell carcinoma. Nature Reviews Disease Primers, 6, Article
92 . https://doi.org/10.1038/s41572-020-00224-3(2017). Performance of the Patient-Reported Outcomes Measurement Information System 29-item profile in rheumatoid arthritis, osteoarthritis, fibromyalgia, and systemic lupus erythematosus. Arthritis Care Research, 69(9), 1312–1321.
(2015). Working memory and executive function decline across normal aging, mild cognitive impairment, and Alzheimer’s disease. Biomedical Research International, 2015, Article
748212 . https://doi.org/10.1155/2015/748212(2001). Cognitive functions of adolescent childhood cancer survivors assessed by event-related potentials. Medical and Pediatric Oncology, 36, 442–450.
(2013). The role of cognitive event-related potentials in executive dysfunction. Kaohsiung Journal of Medical Sciences, 29, 680–686.
(2019). Cancer-related cognitive impairment: An update on state of the art, detection, and management strategies in cancer survivors. Annals of Oncology, 30, 1925–1940.
(2021). Cognitive and neuroimaging markers for preclinical vascular cognitive impairment. Cerebral Circulation; Cognition and Behavior, 2, Article
100029 . https://doi.org/10.1016/j.cccb.2021.100029(2014). An introduction to the event-related potential technique (2nd ed.). MIT Press.
(1987). The short-form McGill pain questionnaire. Pain, 30(2), 191–197.
(2020). Event-related-potential (ERP) markers of traumatic brain injury (TBI) severity and cognitive function – Understanding how the brain works and thinks post TBI. Journal of Systems and Integrative Neuroscience, 6, 1–12.
(2019). Chronic systemic symptoms in head and neck cancer patients. Journal of National Cancer Institute Monographs, 2019(53), Article
lgz004 . https://doi.org/10.1093/jncimonographs/lgz004(2005). [P-001]: Cognitive ERPs as biomarkers and predictors of dementia in mild cognitive impairment (MCI). Alzheimer’s and Dementia, 1(S1), S9. https://doi.org/10.1016/j.jalz.2005.06.063
(2011). Cognitive event-related potentials: biomarkers of synaptic dysfunction across stages of Alzheimer’s disease. Journal of Alzheimer’s Disease, 26(3), 215–228.
(2008). Usefulness of event-related potentials in the assessment of mild cognitive impairment. BMC Neuroscience, 9, Article
107 .(2018). Cognitive impairment associated with cancer. Innovations in Clinical Neuroscience, 15(1–2), 36–44.
(2005). BRIEF-A: Behavior Rating Inventory of Executive Function – Adult version: Professional manual. Psychological Assessment Resources.
(2014). The error processing system in major depressive disorder: cortical phenotypal marker hypothesis. Biological Psychology, 99, 100–114.
(2014). Effects of mindfulness-based cognitive therapy on neurophysiological correlates of performance monitoring in adult attention-deficit/hyperactivity disorder. Clinical Neurophysiology, 125, 1407–1416.
(2014). Modulation of induced frontocentral theta (Fm-theta) event-related (de-)synchronisation dynamics following mindfulness-based cognitive therapy in Major Depressive Disorder. Cognitive Neurodynamics, 8, 373–388.
(2015). Multi-dimensional modulation of α and γ cortical dynamics following mindfulness-based cognitive therapy in major depressive disorder. Cognitive Neurodynamics, 9, 13–29.
(2018). The experimental parameters that affect attentional modulation of the ERP C1 component. Cognitive Neuroscience, 9(1–2), 53–62.
(2009). Event-related potential: An overview. Indian Psychiatry Journal, 18(1), 70–73.
(2012). Psychological disorders, cognitive dysfunction and quality of life in nasopharyngeal carcinoma patients with radiation-induced brain injury. PLoS One, 7(6), Article
e36529 .(2020).
Electrophysiology monitoring . In A. H. B. WuW. F. PeacockEds., Biomarkers for traumatic brain injury (pp. 113–142). Academic Press.(2019). A primer on electroencephalography and event-related potentials for organizational neuroscience. Organizational Research Methods, 22(1), 69–94.
(2014). Epidemiological trends in head and neck cancer and aids in diagnosis. Oral Maxillofacial Surgery Clinics of North America, 26(2), 123–141.
(2000). The visual N1 component as an index of a discrimination process. Psychophysiology, 37, 190–203.
(2018). Event-related potentials during sustained attention and memory tasks: utility as biomarkers for mild cognitive impairment. Alzheimer’s and Dementia: Diagnosis, Assessment, and Disease Monitoring, 10, 452–460.
(2022). Prevalence of cognitive impairment following chemotherapy treatment for breast cancer: A systematic review and meta-analysis. Scientific Reports, 12, Article
2135 . https://doi.org/10.1038/s41598-022-05682-1(2020). Cognitive impairments in breast cancer survivors treated with chemotherapy: A study based on event-related potentials. Cancer Chemotherapy and Pharmacology, 85(1), 61–67. https://doi.org/10.1007/s00280-019-03994-0
(2008). Radiation dose, driving performance, and cognitive function in patients with head and neck cancer. Radiotherapy and Oncology, 87, 304–307.
(2018). Association of neurocognitive deficits with radiotherapy or chemoradiotherapy for patients with head and neck cancer. JAMA Otolaryngology – Head & Neck Surgery, 144(1), 71–79. https://doi.org/10.1001/jamaoto.2017.2235
